ROUNDTABLE: Running AI Inference at Scale

In this roundtable, Nebius will share practical insights and real-world lessons from operating a high-performance, cost-efficient inference platform used by leading AI-native companies.

Attendees will gain a clear understanding of how to optimize inference at scale through techniques such as speculative decoding, prefill acceleration, quantization, and model-level optimizations that dramatically improve latency and throughput.

Nebius will walk participants through proven approaches for reducing cost per token, handling variable workloads, and extracting maximum performance from modern open-source models.